scipopt
If compile SCIP with option of TPI=tny while using whole SCIPsuite tarball, seems it makes crash
Describe the bug There are two environments I tried. Both used python version 3.10.
First one is Windows server 2012 R2. I used anaconda environment, and installed most recent pyscipopt with anaconda.
Also I could not pre-compile with tarball since the OS does not supports the WSL. This means, I used scip which automatically installed with pyscipopt.
In this case, every models I tried, even with almost 20k constraints, was optimized with showing output normally. The output was shown per every 3 to 5 minutes.
I was using model.setParam("display/freq", 20000), and there was no problem in windows environment.
Second one is Ubuntu 20.04 LTS. I used miniconda environment. I downloaded the tarball of full SCIPsuite, and only replaced scip folder inside with main branch of scip. I compiled full SCIPsuite with replaced scip with option TPI=tny. Then added the installed folder as environment variable. Also I installed pyscipopt with master branch using pip. In this case, it does not showed any output - but seems it was solving since the computer made really big noise. However, when noise was stopped, jupyter kernel had timed out, and no results came out.
To Reproduce I cannot upload the experienced model, but I think if you make a model with constraints more than 10k and make similar environment as I wrote, then it can be reproduced.
Expected behavior The output table must be shown.
Screenshots
Sorry that I needed to hide the codes; but as you can see, the cell which is optimizing was started almost half of day ago. However, the CPU spec of linux environment is as this:
And memory spec is as this:
This is almost same with windows environment which shows output very well.
System Skip this because already explained on question.
Additional context These are my theories. First, as I mentioned, I used whole SCIPsuite when compiled. Maybe I needed to compile only scip. Second, it can be a problem which happens only on Ubuntu 20.04. Third, it might be a crash from optional libraries; some optional ones were essential in reality.
https://github.com/scipopt/scip/issues/106#issue-2506886900
Seems first theory was right; rather I should say, SCIP itself has a problem.
Hey @adover134, for some reason this one slipped my mind for a long time, I'm really sorry!!
Is the issue still reproducible with the latest release?
Thank you for checking here, @Joao-Dionisio. However, I am currently not working with SCIP. Also, I gave up the problem I was working on, so I am currently not able to access the computer I was working on... This means I cannot check whether the console output problem is reproducible. I can ask to some other person who can check this if he has time.
Thank you, @adover134!
Out of curiosity, was the project part of some company? If so, consider letting us know, we're always interested in knowing how our software is being used :)
I was using for some research in one of the lab of university, with concern of expanding to co-op on some company. It was cancelled, so I think it is ok to tell the subject. I was working for solve piping plan problem without using graph modelling - only with MILP.